RE: Flash Copy questions

Justin;

IMO the time to IPL a flashcopy "copy" of the production LPAR is mostly
irrelevant.

When you are ready to backup the partition (I assume you are using BRMS)
you run the flashcopy job on the controlling partition,
the controlling partition communicates with both the HMC, your production
LPAR, and external storage sub-system. The job
basically performs a quiesce of the production system (without taking the
production system down) and the external storage creates
a point-in-time snapshot of the production disk. A new LPAR is dynamically
started up/ipl-ed and those disk for the backup lpar are
copied/backedup to to tape/vtl using standard brms system saves.

We have been using this technology for about 4-5 years and basically
perform complete system backups of every day (the largest lpar
is about (8-10TB). Backing up to 3592 (parallell sves to tape) we are
getting about 800-1100 Gb per hour.

as long as the backup LPAR IPL and backup to tape meets your backup window
requirements it can take several hours without bringing down
or impacting your production system.

Good primers for flashcopy are located here:
https://ibm.co/2JBtvb0

Jim

Jim Grant
PDP Group, Inc.
www.pdpgroupinc.com

Confidentiality Notice: This email message is intended only for the
individual or entity to which it is addressed. This email may contain
information that is proprietary or privileged, confidential and exempt
from disclosure under applicable law. If you are not the intended
recipient, you are hereby notified that any dissemination, distribution or
copying of this communication is strictly prohibited. If you received this
email by accident, please notify the sender immediately and destroy this
email and all copies of it.




From: "Justin Taylor" <JUSTIN@xxxxxxxxxxxxx>
To: "Midrange Systems Technical Discussion" <midrange-l@xxxxxxxxxxxx>
Date: 06/14/2018 04:11 PM
Subject: RE: Flash Copy questions
Sent by: "MIDRANGE-L" <midrange-l-bounces@xxxxxxxxxxxx>



Thanks for the comprehensive feedback.


Our BP says that IBMi/SWMA is charged by hardware, so additional
partitions are no extra charge.

We were told the only way to do a Flash Copy snapshot was with an IPL. I
had some concerns about that, in addition to the time required.

The 1TB would drop like a bomb as part of the migration. The plan calls
for 12 drives, in 2 RAID sets with a hot-spare drive in each set. That
would give us 8 drive's worth of space. At double our current growth,
we'd have 72 months' worth of growth. That's factoring in the Flash Copy,
but ignoring the space for the controlling partition.

I expect the overhead of the external storage and the extra partitions
will slightly diminish our performance gains, but the improvement will
still be enormous.





-----Original Message-----
From: Musselman, Paul [mailto:pmusselman@xxxxxxxxxxxxxxxx]
Sent: Thursday, June 14, 2018 1:47 PM
To: Midrange Systems Technical Discussion <midrange-l@xxxxxxxxxxxx>
Subject: RE: Flash Copy questions

Justin--

We're on our 2nd external storage device. First was a V7000 with a Texas
Memory Systems flash appliance tacked on the side. We're now running on a
V9000. The V9000 is 100% flash memory, no disk. This is NOT Flash Copy!
Just very fast storage.

Flash Copy lets you clone your LPAR. This lets you back up things without
bothering the production copy.

1) Partitions. You need the original (all storage space allocated), a
control (all storage space allocated, but not a lot required.), and the
copy ("Thinly Provisioned;" space allocated as needed). IIRC, the control
can run on the original (but don't hold me to that). Yes, this is 3
copies of IBM i, 3x the cost. But consider the benefits.
a) You can run an Option/21 backup of your main system
whenever you want to. We run one every night! Until we could make a
Flash Copy we only ran an Option/21 every 3 years, when we upgraded the OS
or the hardware!
b) You have a perfect copy of your production for any
stray testing you might want to do. Yes, you should have a 'permanent'
test environment, but you can never have too many!
c) Migrating from one external storage system to a newer
system is a piece of cake! Make sure your PTFs are in order, then roll in
the new hardware, the business partner's technical wizard works some magic
to handle 2 major things:
i) your data is migrated from the old
storage to the new storage while everything keeps running!
ii) Down time is about an hour to switch
the load source to the new hardware, and you're up and running!
d) Check and see what migration tools are available from
your 720 to external storage. IIRC there was some magic possible from
internal storage to external storage. I don't think we had to
unload/reload to migrate to the V7000.

1a) Thinly Provisioned storage for the Flash Copy LPARs means that you
define a block of storage for the Flash Copy just as big as your
production LPAR, but you don't allocate it until you need it. And the
storage will allocate just enough to hold the parts of storage that have
changed-- not an entire file, and not an entire record, but just the
blocks of memory that contain the parts of the records that have changed.
As the Flash Copy 'ages,' it gradually grows as things change on either
the original or the copy. See #2) below. When you tell the storage to
shut down the Flash Copy, all that space goes back into 'unused'
territory.

2a) You said, " Flash Copy requires 3 partitions (control, production &
target). During an IPL, the production partition can be imaged to the
target partition."
Your original LPAR should behave just as it always has. The only thing
the Flash Copy does is clone the original (when you need a copy). You'll
be getting an HMC (Hardware Management Console). This will control the
creation of a Flash Copy whenever you want. We're using the Full System
Flash Copy Toolkit (or whatever the current name is). The Toolkit allows
you to configure the cloning to run automatically. Our daily backup runs
automatically, creating the Flash Copy, running the backup, and ejecting
the tapes.

Imaging the production partition to the target partition at IPL time can
be an advantage-- you're creating the Flash Copy when the original system
is 'quiet.' So there are no dangling commitment boundaries and no
half-updated transactions to worry about. We've accepted the chance of
things stuck in-transit because we don't want to go to Restricted State
every day just to run backups. Generally, we haven't had issues. But
Murphy is probably waiting for an opportune moment.

2b) Creating the Flash Copy takes about 2 minutes. During that time, Main
Storage for the LPAR is flushed to disk, and disk operations to the
original LPAR are suspended. Depending on how large Main Storage is the
time required will vary. Users will not usually notice this time, unless
they're actively updating files. Once the system has been Quiesced, the
actual Flash Copy is over in seconds. Then disk operations to the
original LPAR resume, picking up where they left off with no glitches.
This is probably best done at a time when users are not pounding away at
the system!

2c) IPL time of the Flash Copy LPAR is not much different than the
original LPAR. There is some overhead-- especially since the copy is
thinly provisioned. The Flash Copy creates a table to track storage space
between the original and the copy. You want to change to original, the
storage clones that data so the copy is 'untouched.' You want to change
the copy, the storage clones the data so you don't damage the original.
This copying adds some overhead. But unless you're very picky, it
shouldn't be a major concern.

3) TANSTAAFL. We migrated to the V7000 blindfolded. Our business partner
at the time was just as blindfolded as we were. We didn't know what to
expect. We've since 'upgraded' business partners!
There is overhead with external storage. The iSeries hardware has to talk
to the external storage to request data. No more disks-just-down-the-bus
to retrieve data from. But the external storage is fast.

We're currently running 3 production LPARs. We have configured Flash Copy
for two of them. We are planning on migrating some external storage from
our network group into the V9000, but that's a future project. So we're
swimming in space at the moment. Having extra space, when we migrated our
systems to the current hardware we never added space to the production
LPARs. But we can add more just by contacting our business partner's tech
wiz to allocate more space. Just takes an hour or so.

It looks like you have 19 GB 'unused' at the moment. How fast does your
data grow? It's probably cheaper to add more storage space during the
conversion than to add it later. And avoids the fun of updating the
hardware later! An extra TB for image-type files sounds like fun-- If you
don't allocate it all at once the 'unused' space becomes available for
Thinly Provisioned Flash Copy space.

Since you're jumping from a 720 to a 914, any overhead from external
storage will be 'hidden' by the faster hardware. You may not see 100% of
the speed upgrade you were expecting, but overall everything will be
faster.

Paul E Musselman
PaulMmn@xxxxxxxxxxxxx